About This Project
The project primarily focuses on identifying specific acoustic signatures that trigger responses from congeneric but phylogenetically distant babbler species. Secondly, it utilizes CNN-based AI models to detect species that form associations in the acoustic space during dawn chorus, serving as a captivating manifestation of intraspecies communication within complex tropical environments while emphasizing the broader context of whole soundscapes to assess biodiversity.
Ask the ScientistsJoin The Discussion
What is the context of this research?
Studies have shown that birds use the information encoded in alarm signals of other species in mixed-species flocks. Social learning through a process of acoustic association can help them in recognizing unknown alarm calls of heterospecific species and increase their survival chances against predators. But social species of babblers like Jungle Babbler and Large Gray Babbler respond even to non-alarm signals of one another. Why do they recognize these calls? They aren't even closely related unlike these wren species. The second part of this project aims to look at soundscapes as a whole to understand patterns of diversity during dawn chorus which seems to be getting quieter and less diverse.
What is the significance of this project?
There should be shared acoustic signatures that foster responses between distantly related babblers in both shared & separate environments and the discovery of those challenges our understanding of interspecies communication, offering new perspectives on the evolution of vocalization in birds. If animals from different species use similar signals to convey information, they might have evolved similar ways of communicating with each other. This convergence might be driven by shared ecological factors, revealing ways in which diverse species adapt to their environments. Furthermore, analyzing soundscapes can act as a valuable proxy for assessing biodiversity. We have observed that they vary with the seasons and across different locations, impacting dawn chorus diversity.
What are the goals of the project?
For my primary objective, vocalizations of both species are collected & multiple acoustic features are extracted from the generated spectrograms. Deep learning models are trained on these features to build a classifier. Model's performance is analyzed to understand if it effectively distinguishes between the vocalizations of two species. Misclassifications might suggest sharing of acoustic features. Secondly, I am collecting dawn chorus recordings. A species detection algorithm based on CNN/RCNN, will be employed to identify avian species vocalizing in the same acoustic space & time as manual annotations are time consuming. Next, we look for correlations between soundscape variations & acoustic space use. This will help us in looking at how dawn chorus is changing spatially & temporally.
Passive acoustic recorders like SongMeter mini provides a non-invasive approach to monitor biodiversity. These can collect huge amount of data in difficult terrains without disturbing the fauna. It has weatherproof design suitable for any environment. Batteries and cards are essential for the functioning of recorders and storage of data with few extra as backups. Assessing remote areas is tough because of poor public transport so personally hired vehicles would be needed for the deployment and undeployment of recorders on a periodic basis.
About one and half year long project. We already have some data but we need to collect more dawn chorus recordings from different habitats to build generalised and robust species detection and classification models. 6 months would be spend on data collection and then we will do data preprocessing and analysis. We have done preliminary soundscape analysis - looking at spatio-temporal differences in soundscapes of different habitat types and have obtained promising results.
Nov 20, 2023
Dec 15, 2023
Finish soundscape analysis
Dec 22, 2023
May 15, 2024
Finish data collection for dawn chorus and babbler vocalizations from multiple contexts.
Jul 15, 2024
Data Preprocessing and annotations for model training.
Meet the Team
Team Member 1: Dr. Manjari Jain (co-PI) [Associate Professor of Behavior Ecology, IISER Mohali]
Team Member 2: Dr. Dan Stowell [Associate Professor of AI & Biodiversity, Tilburg University]
Dr. Manjari Jain, my PhD supervisor, is a distinguished biologist with a specialization in behavioral ecology and acoustic communication. Meanwhile, Dr. Dan Stowell, an authority in computational bioacoustics, will serve as my project advisor, providing invaluable guidance on classifier algorithms.
I am a PhD student trying to merge bioacoustics and artificial intelligence to obtain something fruitful for ecology. My journey in science has been quite of a roller coaster ride. After finishing my masters specializing in high energy physics, I shifted gears to the fascinating field of ecology. My motivation for this was nothing but the simple reason of looking at birds. A core-year course on behavior ecology developed my interest in birds and I took bird watching as a hobby. After finishing my masters I decided to convert this hobby into something more meaningful.
Nothing posted yet.
The project has already started and recordings are being collected and annotated. We already have done some pilot analyses for all three objectives. However we lost our recorders (audiomoth - cheap but unreliable) to weather conditions and living in a poor country like India, we can't afford to buy new ones. So the project is stuck.
- $4,905Total Donations
- $2,452.50Average Donation